For a good three years, ChatGPT was the only restaurant in town.

Now there's a food court. ChatGPT's share of generative AI web traffic fell from 73.3% to 55.5% in the year to August 2026, while Gemini doubled its slice and Claude went from rounding error to nearly one visit in ten.

So is ChatGPT about to become the BlackBerry of AI: much loved, used by everyone, then suddenly nothing but a gameshow answer? Here's what the 2026 data says 2027 will bring…

Where ChatGPT starts in 2027

Any forecast needs a baseline. ChatGPT heads into 2027 bigger than ever in absolute terms, and smaller as a share of the market.

That combination of huge reach, fast revenue growth, and heavy spending shapes everything that follows.

Washington AI companies putting agents into production
Washington pairs the federal government's agentic ambitions with enterprises shipping multi-agent systems at scale. Here is who builds what around the Beltway.

Prediction 1: ChatGPT becomes a work platform first

The biggest shift in 2027 will happen inside the office. OpenAI president Greg Brockman said in November 2026 that businesses now bring in most of OpenAI's revenue, building on a May 2026 plan to fold ChatGPT and Codex into one agentic platform.

The groundwork is already visible. ChatGPT and Codex usage limits are set to merge in December 2026. OpenAI reports over 35 million weekly ChatGPT Work and Codex users, and says its products reach 2.5 million businesses.

💡
Expect the 2027 version of ChatGPT to spend less time answering questions and more time completing tasks: writing and shipping code, running multi-step research and acting across connected apps. The chat box stays, but it becomes the front door to a team of agents.

Prediction 2: ads at scale, and ChatGPT goes public

The free tier will pay its way through advertising. ChatGPT Ads reached a $1 billion annualized run rate in under 200 days and expanded to 89 countries on October 21, 2026, across the Free and Go plans. OpenAI has reportedly told investors to expect $11 billion in ad revenue in 2027.

2027 is also the reported IPO year. Brockman said he believes OpenAI will be ready to list in 2027, and CFO Sarah Friar reportedly told staff it will be a public company that year. No date has been set.

A listing changes the product. Public markets reward predictable revenue, so expect sharper monetization: more ad formats, tighter free-tier limits and harder pushes toward paid plans. A public S-1 will also put those operating losses under a much brighter light.

I agent audit: 18 of 25 cognitive modules sat idle
Theory of mind, causal reasoning, global workspace: a recent arXiv audit found most named modules in one LLM agent were decoration. Rewiring two of them took a planner to a perfect score.

Prediction 3: ChatGPT leaves the screen

OpenAI's first consumer hardware is expected in 2027. The company said in a February 2026 court filing that the device will not ship before the end of February 2027.

Bloomberg reported in July 2026 that it will be a screenless smart speaker designed with Jony Ive's LoveFrom. It will use a camera, sensors, and an advanced version of OpenAI's GPT-Live voice model to read its surroundings.

This matters for the BlackBerry question. Google's edge is distribution through Android, Chrome and Search; a device is OpenAI's attempt to own a channel of its own.

💡
Hardware is also the one roadmap item OpenAI has already slipped once, so treat the date as a target. And yes, a camera that never blinks will need a very good privacy story.

Prediction 4: still the biggest, with a shrinking slice

Expect ChatGPT to end 2027 as the largest AI assistant, with a smaller share of the market than it holds today. The trend lines all point that way.

Share of generative AI web visits

August 2025

August 2026

ChatGPT

73.3%

55.5%

Gemini

12.9%

25.6%

Claude

1.9%

9.3%

[Source: Similarweb. Web visits only; app usage is excluded, which undercounts Gemini on Android.]

Gemini's distribution. The Gemini app passed 1 billion monthly users in August 2026, and Google keeps building it into Android, Chrome, Workspace and Search. By Sensor Tower's measure, ChatGPT's share of unique AI assistant users already fell below 50% in March 2026.

25 AI researchers to follow in 2027
Discover the researchers before the roadmap catches up…

The enterprise contest.

Menlo Ventures' November 2026 report puts Anthropic at 55% of enterprise LLM API spend, up from 40% a year earlier, against 31% for OpenAI. Menlo is an Anthropic investor, so read it alongside other data.

The Ramp AI Index shows 43.8% of US businesses paying Anthropic in August 2026 versus 39.8% paying OpenAI. OpenAI's Codex push is aimed squarely at this gap.

Cheap open models. DeepSeek took 24.4% of text requests on OpenRouter in the week of September 21, 2026, ahead of Google and OpenAI.

Price pressure at the API layer will keep growing.

💡
The defining 2027 behavior will be multi-homing. Ramp's figures imply at least 27% of its sampled businesses already pay both Anthropic and OpenAI. Users are assembling a portfolio of assistants, one per job.

The BlackBerry test for 2027

BlackBerry defined the smartphone for professionals, then lost to Apple and Google, who owned the operating system and the app ecosystem. Its phone sales kept rising for a few years after the iPhone launched, even as its share collapsed. So rising user numbers alone prove nothing.

The parallels are real: a category creator, a rival with platform-level distribution in Google, and an enterprise base under pressure.

Here are the signals that would mean ChatGPT is following the same path in 2027:

  • Weekly users stall or fall. Share can slip for years without harm; a flat or shrinking absolute user count would be a genuine red flag.
  • The agent platform fails to win developers. If the ChatGPT and Codex merger fails to close the enterprise gap, OpenAI loses the customers who now pay most of its bills. Agent experience will decide a lot of that.
  • Ads erode trust. If users start to treat ChatGPT answers like sponsored search results, the habit that powers its lead weakens fast.
  • The device flops or slips again. Without its own channel, OpenAI stays dependent on platforms owned by its biggest rivals.
💡
The analogy breaks on one key point. BlackBerry missed the platform shift; OpenAI is driving it, shipping frontier models every few months. Its real risk is commoditization: when OpenAI, Google, and Anthropic models all answer the same question well, the logo on the tab matters less.

A closer comparison may be Google Search in the late 2000s: the default that stayed the default while specialists picked off valuable jobs one by one.


What AI teams should do before 2027

Whatever ChatGPT becomes, betting your stack or your traffic on a single assistant is the riskiest option on the table.

The teams that win in 2027 will treat AI assistants the way they treat cloud providers: useful, powerful and never irreplaceable.

9 AI predictions for 2027 for your planning deck
Your 2027 planning deck needs sharper assumptions. These nine analyst predictions cover agent governance, power limits, regulation, and physical AI…

The verdict: a different kind of king

💡
ChatGPT in 2027 will most likely be a listed company's flagship product, an agent platform for work, an ad-supported consumer app and a voice in a new device. It will still have more users than any rival.

What it will lose is the illusion of being the only option. Gemini will keep winning on distribution, enterprises will keep spreading spend across vendors, and open models will keep pushing API prices down.

That is a long way from BlackBerry. The real danger is quieter: a slow slide from default to one tab among three. Watch weekly users, enterprise revenue, and how people react to ads. If those hold through 2027, ChatGPT keeps the crown, just with a smaller kingdom.